So, can someone confirm RGB led works over i2c? This is the only piece which seems missing at this point 
Update: finally cracked it. RGB channels are driven by LP5562. And it has upstream driver. Wiring it up
Another update: and this one could probably be upstreamed on its own
- Ported edma_v2 approach to _v1. Basically using 4 CPUs and related structures vs just 1
- Performance numbers are very promising: 2.6gbps went up to 4.6gbps - with a 5G+10g clients. Will retest tomorrow with 2 10G clients
- Added checksum offloads too - another .2G bump
Update: and a major performance win: wan10g <-> lan10g throughput - hardware offloaded end to end (AQR - IPQ - RTL - ARQ) with ~0% CPU load 
┌───────────────────┬─────────────┬─────────────┐
│ Test │ Upload │ Download │
├───────────────────┼─────────────┼─────────────┤
│ TCP single stream │ 9.38 Gbit/s │ 9.39 Gbit/s │
├───────────────────┼─────────────┼─────────────┤
│ TCP ×4 streams │ 9.39 Gbit/s │ 9.39 Gbit/s │
└───────────────────┴─────────────┴─────────────┘
and more interesting numbers
┌──────────────────────────────────────────┬─────────────┬──────┬────────────────────────────────────────────────────────────────────┐
│ test │ wired │ wifi │ verdict │
├──────────────────────────────────────────┼─────────────┼──────┼────────────────────────────────────────────────────────────────────┤
│ wired up alone │ 9.39 │ — │ baseline │
├──────────────────────────────────────────┼─────────────┼──────┼────────────────────────────────────────────────────────────────────┤
│ wired up + wifi down (the original case) │ 9.38 │ 1.49 │ no drop at all, even with 50% more wifi load than the original run │
├──────────────────────────────────────────┼─────────────┼──────┼────────────────────────────────────────────────────────────────────┤
│ wired up + wired down (both silicon) │ 9.36 / 9.09 │ — │ 18.4 Gbit/s aggregate NAT in the PPE │
├──────────────────────────────────────────┼─────────────┼──────┼────────────────────────────────────────────────────────────────────┤
│ wired up + wifi up (same direction) │ 7.11 │ 1.52 │ real contention │
└──────────────────────────────────────────┴─────────────┴──────┴────────────────────────────────────────────────────────────────────┘
┌────────────────────┬───────────────┬─────────────┬────────────┐
│ 6 GHz HE160 (ch33) │ single stream │ P4 │ router CPU │
├────────────────────┼───────────────┼─────────────┼────────────┤
│ upload │ 1.54 Gbit/s │ 1.56 Gbit/s │ ~28% │
├────────────────────┼───────────────┼─────────────┼────────────┤
│ download │ 1.35 Gbit/s │ 1.54 Gbit/s │ ~25–35% │
└────────────────────┴───────────────┴─────────────┴────────────┘
@a_guy I assume you have access to the device on the latest firmware etc? I have a sibling device that runs the same code it appears that I bribed into giving up the goods. No real use for it but it seems the decrypted rootfs and all that isn't very easy to obtain. CR200A running 3.5.0.10 (Qualcomm IPQ8072A / SDX LEMUR). If anything from it would be of help to you in your quest I'd be happy to connect offline. I would guess that the methods work on its family pretty universally..